Microsoft is transitioning GitHub Copilot to a token-based billing model, marking a significant shift from its current requests-based system and prompting changes to pricing, usage limits, and model availability across subscription tiers.
Internal documents reviewed by technology journalists reveal that the company plans to suspend new signups for individual paid tiers of GitHub Copilot while implementing the new billing approach, which charges users based on the actual number of tokens processed rather than discrete interactions with the AI assistant.
The shift comes as the weekly cost of running GitHub Copilot has nearly doubled since the start of the year, according to the same internal sources, making the current subsidized model unsustainable at scale.
Under the existing requests-based system, GitHub Copilot Pro subscribers receive 300 monthly interactions, while Pro+ users receive 1,500. However, each request can consume varying numbers of tokens depending on the AI model used, with more advanced models requiring significantly more computational resources per interaction.
Token-based billing will instead charge users for both input tokens (prompts and context provided to the model) and output tokens (the model’s responses, including any reasoning steps), aligning costs more closely with actual compute usage. For reference, models like Anthropic’s Claude Opus 4.7 are priced at $5 per million input tokens and $25 per million output tokens in external API offerings.
As part of the transition, Microsoft will remove access to the Anthropic Opus family of models from the GitHub Copilot Pro tier, reserving them exclusively for the more expensive Pro+ subscription. This change reflects the higher computational cost associated with running advanced models.
The company also plans to tighten rate limits on certain GitHub Copilot Business and Enterprise accounts and will discontinue free trials for paid individual plans as part of broader efforts to curb misuse and manage infrastructure expenses.
Microsoft has confirmed some of these details publicly, noting that the move to token-based billing will result in higher costs for most individual users compared to the current flat-rate subscriptions, which have not increased in price despite rising operational expenses.
While no exact timeline for the full rollout has been provided in the verified sources, the company indicates that the formal announcement and implementation are imminent, with changes expected to take effect in the near term as Microsoft seeks to align pricing with actual resource consumption.
The shift underscores broader trends in AI infrastructure economics, where providers are moving away from simplified usage metrics toward granular, consumption-based models that reflect the true cost of large language model deployment at scale.
